Characterization of Vision-Aided Indoor Localization and Landmark Routing

GPS is the premier method of localization and way finding in outdoor environments. Indoor environments prevent GPS from functioning properly or at all. RF-based solutions have been proposed using varieties of radio triangulation and spectral fingerprinting. Such technically attractive methods have not seen widespread adoption in places where the demand is greatest: supermarkets, shopping malls, airports, etc. This may be due to the obstacles of funding, installing, or accessing sufficient wireless infrastructure for triangulation, as well as the scalability challenge in site-by-site fingerprinting. We present a framework for characterizing indoor environments. This leads to a simple, practical approach to indoor localization and wayfinding that takes advantage of visibility relationships, limited user-input, publicly available online floorplans, and lightweight processing on a mobile phone. On-site infrastructure and site-surveys are avoided. Based on this framework we find that localization precision improves dramatically as the population of landmarks grows, though few landmarks are needed during a given input cycle. We also find that indoor wayfinding is unexpectedly insensitive to hop-count, but critically dependent on path connectedness.